Search

This paper presents a neural network-based method to generate multiple images with different exposures from a single input low dynamic range (LDR) image for improved high dynamic range (HDR) imaging. The proposed algorithm consists of three steps: 1) 2-D histogram estimation; 2) neural network-based LDR images estimation; and 3) generation of an optimal set of differently exposed images. The proposed method first generates image features by estimating a patched-based 2-D histogram. The extracted features are used in an input layer of the neural network, which plays a role to select an optimal set of LDR images. A set of LDR images is generated using a curvature-based contrast enhancement method. Experimental results show that the proposed method can generate an optimal set of LDR images using neural network and provide improved HDR images. In addition, the proposed method can be implemented as a preprocessing step in most existing HDR frameworks. The proposed HDR approach is considered as a single-input method that gives almost the same performance to multiple image-based HDR method. [ABSTRACT FROM PUBLISHER]

The problems of the neural network-based nonuniformity correction algorithm for infrared focal plane arrays mainly concern slow convergence speed and ghosting artifacts. In general, the more stringent the inhibition of ghosting, the slower the convergence speed. The factors that affect these two problems are the estimated desired image and the learning rate. In this paper, we propose a learning rate rule that combines adaptive threshold edge detection and a temporal gate. Through the noise estimation algorithm, the adaptive spatial threshold is related to the residual nonuniformity noise in the corrected image. The proposed learning rate is used to effectively and stably suppress ghosting artifacts without slowing down the convergence speed. The performance of the proposed technique was thoroughly studied with infrared image sequences with both simulated nonuniformity and real nonuniformity. The results show that the deghosting performance of the proposed method is superior to that of other neural network-based nonuniformity correction algorithms and that the convergence speed is equivalent to the tested deghosting methods. [ABSTRACT FROM AUTHOR]

Changing the resolution of digital images and video is needed image processing systems. In this paper, we present nonlinear interpolation schemes for still image resolution enhancement. The proposed neural network interpolation method is based on wavelet reconstruction. With the wavelet decomposition, the image signals can be divided into several time--frequency portions. In this work, the wavelet decomposition signal is used to train the neural networks. The pixels in the low-resolution image are used as the input signal of the neural network to estimate all the wavelet sub-images of the corresponding high-resolution image. The image of increased resolution is finally produced by the synthesis procedure of wavelet transform. In the simulation, the proposed method obtains much better performance than other traditional methods. Moreover, the easy implementation and high flexibility of the proposed algorithm also make it applicable to various other related problems. [ABSTRACT FROM AUTHOR]

In this paper, Levenberg-Marquardt feedforward MLP neural network (LMFFNN) was proposed to classify cervical cell images obtained from 100 patients including healthy, low-grade intraepithelial squamous lesion and high-grade intraepithelial squamous lesion cases. This neural network along with extracted cell image features is a new model for cervical cell image classification. The semiautomated cervical cancer diagnosis system is composed of two phases: image preprocessing/processing and feedforward MLP neural network. In the first stage, image preprocessing is done to reduce the existing noises without lowering the resolution. After that, image processing algorithms were applied to manually cropped cell images to achieve a linear plot which includes real components, were used as LMFFNN inputs for classification of cervical cell images. Based on the results, cervical cell images were classified successfully with 100 % correct classification rate using the proposed method. Moreover, the rates of sensitivity and specificity were calculated as 100 % using LMFFNN method. It was shown there was a good agreement between the expert decision and values gained from the ANN model. [ABSTRACT FROM AUTHOR]

People counting is a very important problem in visual surveillance. An accurate and real time estimation of people in a crowded place can provide valuable information. Here video is given as input and outputs the average number of people passing over the video. The video input is separated to number of frames and some processing steps are performed on background subtraction results to estimate the number of people in a complicated scene. Foreground pixel extraction is done with ViBe (VIsual Background Extractor) algorithm. The extracted foreground image's pixels count is calculated and gives as input to the neural network. In learning phase, the people count is calculated manually with test dataset and while testing phase remaining test cases are tested by adjusting weight parameters to obtain relative to the target result. [ABSTRACT FROM AUTHOR]

In this paper, we present a new neural network (NN) for three-dimensional (3D) shape reconstruction. This NN provides an analytic mapping of an initial 3D polyhedral model into its projection depth images. Through this analytic mapping, the NN can analytically refine vertices position of the model using error back-propagation learning. This learning is based on shape-from-shading (SFS) depth maps taken from multiple views. The depth maps are obtained by Tsai–Shah SFS algorithm. They are considered as partial 3D shapes of the object to be reconstructed. The task is to reconstruct an accurate and complete representation of a given object relying only on a limited number of views and erroneous SFS depth maps. Through hierarchical reconstruction and annealing reinforcement strategies, our reconstruction system gives more exact and stable results. In addition, it corrects and smoothly fuses the erroneous SFS depth maps. The implementation of this neural network algorithm used in this paper is available at http://kumazawa-www.cs.titech.ac.jp/~fanany/MV-SPRNN/mv-sprnn.html. [Copyright &y& Elsevier]

Many existing scene-adaptive nonuniformity correction (NUC) methods suffer from slow convergence rate together with ghosting effects. In this paper, an improved NUC algorithm based on total variation penalized neural network regression is presented. Our work mainly focuses on solving the overfitting problem in least mean square (LMS) regression of traditional neural network NUC methods, which is realized by employing a total variation penalty in the cost function and redesigning the processing architecture. Moreover, an adaptive gated learning rate is presented to further reduce the ghosting artifacts and guarantee fast convergence. The performance of the proposed algorithm is comprehensively investigated with artificially corrupted test sequences and real infrared image sequences, respectively. Experimental results show that the proposed algorithm can effectively accelerate the convergence speed, suppress ghosting artifacts, and promote correction precision. [ABSTRACT FROM AUTHOR]

Abstract: Depth map is used for recovery of three-dimensional structure of the object which is required in many high level vision applications. In this paper, we present a new algorithm for the estimation of depth map for three-dimensional shape recovery. This algorithm is based on Fuzzy-Neural approach using shape from focus (SFF). A Fuzzy Inference System (FIS) is designed for the calculation of the depth map and an initial set of membership functions and fuzzy rules are proposed. Then Neural Network is used to train the FIS. The training is done using back propagation algorithm in combination with the least squares method. Hence, a new set of input membership functions are generated while discarding the initial ones. Lastly, the trained FIS is used to obtain final depth map. The results are compared with five other methods including the traditional SFF method and the Focused Image Surface SFF method (FISM). Six different types of objects are used for testing the proposed algorithm. [ABSTRACT FROM AUTHOR]

This paper demonstrates the feasibility of the automation of forensic hair analysis and comparison task using neural network explanation systems (NNESs). Our system takes as input microscopic images of two hairs and produces a classification decision as to whether or not the hairs came from the same person. Hair images were captured using a NEXTDimension™ video board in a NEXTDimension™ color turbo computer, connected to a video camera. Image processing was done on an SGI indigo workstation. Each image is segmented into a number of pieces appropriate for classification of different features. A variety of image processing techniques are used to enhance this information. Use of wavelet analysis and the Haralick texture algorithm to pre-process data has allowed us to compress large amounts of data into smaller, yet representative data. Neural networks are then used for feature classification. Finally, statistical tests determine the degree of match between the resulting collection of hair feature vectors. An important issue in automation of any task in criminal investigations is the reliability and understandability of the resulting system. To address this concern, we have developed methods to facilitate explanation of neural network’s behavior using a decision tree. The system was able to achieve a performance of 83% hair match accuracy, using 5 of the 21 morphological characteristics used by experts. This shows promise for the usefulness of a fuller scale system. While an automated system would not replace the expert, it would make the task easier by providing a means for pre-processing the large amount of data with which the expert must contend. [Copyright &y& Elsevier]

AI trained neural network model is used to generate compensated image to correct the display panel nonuniformity, with camera captured input images of display panel as the inputs for model training. Usage cases are illustrated with simulated 2D BLU for LCD and MiniLED tiled displays. [ABSTRACT FROM AUTHOR]